The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local Python scripts (harness/scripts/health-check.py, harness/scripts/okf-check.py, and harness/scripts/wiki-health.py) to perform its diagnostic functions. These scripts are part of the skill's execution harness and are used for file hashing and version comparison.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests and processes data from external configuration files (.template-manifest.json) and a remote version file. However, this is part of the core diagnostic functionality, and the platform-level disable-model-invocation: true setting significantly limits the risk of the agent being manipulated by malicious content within those files.
Ingestion points: Reads .template-manifest.json, harness/version.json, and fetches version.json from the remote Rheinmir/setup repository.
Boundary markers: None specified in the documentation.
Capability inventory: The skill can execute three specific Python scripts located within the project harness.
Sanitization: The documentation does not specify sanitization for the manifest or version data, as the scripts primarily perform deterministic hash and version comparisons.

health-check